Research

Germline-somatic Continuum in children with Acute Lymphoblastic Leukaemia
We intend to map and describe correlations of germline and somatic events in individual patients with gene expression and molecular signalling, and their association to risk assessment and patient stratification by detecting germline, tumour and somatic genetic variants, structural variants, gene fusions and gene expression. 

Rare Clinical case: Philadelphia + ALL sisters (Ph+ ALL N=2 sisters + 3 unrelated Ph+ ALL patients)
Based on Germline DNA (WGS), Tumour DNA (WGS) and Tumour RNA (RNAseq), we explore the tumour and germline mutational landscape of the sisters, and look for driver candidate events in Ph+ ALL.

Prediction of time to insulin requirement in patients with type 2 diabetes using artificial intelligence
The rate of progression in type 2 diabetes is heterogenous and hard to determine for individual patients. This study aims to understand how predictable time to insulin is for the individual patient using information from electronic medical records and genotype using machine learning approaches.

Asparaginase-associated pancreatitis cases in childhood acute lymphoblastic leukemia
Asparaginase-associated pancreatitis frequently affects children treated for acute lymphoblastic leukemia causing severe acute and persisting complications. Known risk factors such as asparaginase dosing, older age and single nucleotide polymorphisms (SNPs) have insufficient odds ratios to allow personalized asparaginase therapy. In this study, we integrate clinical information and SNP genotype in machine learning models to build predictive classifiers of individual toxicity risk.

Machine learning models to guide thiopurine/methotrexate therapy of acute lymphoblastic leukemia
1-2 years of thiopurine/methotrexate maintenance therapy (MT) is one of the most important treatment phases of childhood acute lymphoblastic leukemia without which 40% of all patients will develop leukemic relapse. Traditionally, MT have been adjusted to obtain a preset degree of myelotoxicity, but this is confounded by wide natural variations in blood counts. Thus, patients are currently seen at 1-2 weeks intervals during MT to titrate therapy to the right dosage. Objectives of the present study: To reach a deeper understanding of interactions of these parameters in order to facilitate drug dosing with two aims: (i) more rapidly reach the individual patient’s target dose AND with fewer outpatient visits, and (ii) ultimately reduce the risk of relapse due to insufficient drug exposure.

Polygenic Risk Scores in multiple population-based cohort studies of childhood cancer risk
Due to the pleiotropic effects and the presence of well-established and validated PRSs on genetic predisposition of adult cancers, we hypothesize that PRSs from adult cancer patients, may additionally be associated with risk of childhood cancers such as acute lymphoblastic leukemia (ALL) or medulloblastomas which represents the most common cancers in children.

Fluctuation modelling of longitudinal patient data
In biological research many variables contain temporal patterns, which makes them time-dependent. In many of these biological processes it is important to understand the pattern to represent this information in models linking association or prediction. The group works with large cohorts such as the UK biobank to develop measurements for volatility, fluctuation and variation in electronic health record data.